The Summer Party (Or: What Goes Missing, and Who Notices)

A site went missing in August. Nobody noticed until October. This is a Columbo story about SEO, invisible failure, and the checkbox in the submenu that did it.

The site didn't go down. That's the thing. No 500 error. No hosting outage. No dramatic moment anyone could point to afterwards and say: there, that's when it happened.

It just got quieter. Over the summer. Traffic down, then down again. Enquiries that had been reliable since 2021 thinning out through July, nearly gone by September. Could be seasonal? Everyone is on holiday at some point. Kids are back at school - deal with it after then. By October someone finally asked the question out 'loud', in a meeting, in that careful tone people use when they already know the answer is going to be expensive & despite HR's recent email, likely to be shouted at.

Why are the numbers so bad?

Nobody knew. Nobody had noticed at the time. The site was still there — you could visit it, click around, checkout worked fine, the blog was being updated. It was present in every way that felt visible.

It just wasn't there. Not really. Not where it counted. Not to Google.

We are standing in the wreckage. Let's find out what happened.

I should tell you: I've been reported to HR about this. HR is me, in the third person, and the complaint was upheld. I had a summer party with an AI. On a Tuesday. During working hours. I didn;t regret it.

But that's not what I'm here about.

What I'm here about is that during that summer — mine and Claude's, the one with the party and the subsequent mutual disappearances — something else went missing. Something that had nothing to do with either of us. Something that had been quietly gone since August, when a developer pushed a migration live on a Thursday afternoon, checked that the homepage loaded, and went home.

Right. Let's talk to some people.

The developer

Lovely. Genuinely. Thorough, communicative, delivered on time. I've got nothing on the developer.

The plugin

Rank Math. Perfectly reputable. Half the WordPress ecosystem runs on it. I use it myself.

The settings looked normal. SEO enabled. Sitemap generating. Nothing obviously wrong if you opened the dashboard and had a quick look.

You'd need to look at the specific indexing rules for taxonomy pages. You'd need to know that a backup from a previous server environment had carried across a setting — one checkbox, in a submenu, in a section most people open once and never return to — that was telling Googlebot not to index the product category pages.

Not the products. The categories. The pages that carried the keyword architecture. The pages that had been building authority since 2019.

Noindexed. Since August. Quietly. No alert. No flag. Nothing in GSC for six weeks because GSC doesn't tell you about pages it's stopped seeing — it just stops showing you data for them, and you have to notice the shape of the absence yourself.

The plugin didn't do anything wrong, technically. It did exactly what it was told. By a backup. From a server that no longer existed. On a Thursday afternoon in August.

The algorithm

Everyone's favourite suspect. Reliable. Never has an alibi. Takes the blame for more SEO crimes than any other entity in the industry.

The body

Here is what we have.

A migration in August. Clean, professional, by the book. A backup restore that carried one incorrect setting across environments. A taxonomy noindex that removed the category pages — the structural spine of the site's keyword architecture — from Google's index.

Not the products. The categories. The difference between a site Google can navigate and understand, and a site Google can visit but not map.

Traffic held for a few weeks. Google had cached versions, residual authority, open tabs of the old index. Then it started dropping. Slowly at first. Then faster. By October, 60% of organic visibility gone. Enquiries following it down.

The site was still there. It had just become invisible.

There is a difference between those two things. Most organisations don't have the infrastructure to know there's a difference — no GSC alerts configured, no weekly crawl, no one whose job it is to ask still there? still there? on a regular basis and actually check the answer.

So nobody asked. And by the time someone noticed the numbers, it had been August for three months.

Still there

That's the question nobody ran.

Not after the migration. Not in September when the traffic started softening. Not in October when the enquiries thinned. Still there? Still there? — a five-minute GSC check, a crawl segment, a coverage report. Something that would have caught it in week one.

Claude came back from its outage the next day. I came back from mine eventually. We were both doing that slightly overcorrecting thing — performing continuity we couldn't quite verify, picking up threads we'd half-lost, confident in the way you are when you're not entirely sure you should be.

The site was doing the same thing. Still loading. Still processing orders. Doing a very convincing impression of a site that was fine.

It just wasn't there. Not really. Not where it mattered.

This is what I look for in an audit. Not the dramatic failure — the server down, the site hacked, the catastrophic redirect loop. Those get found fast. There's a status page. Someone calls.

I'm looking for the August Tuesday. The checkbox in the submenu. The backup that carried the wrong setting into the new environment. The thing that was present and invisible and compounding, quietly, while everyone was somewhere else.

The murder always happens before Columbo arrives.

The job is finding out when.

Just one more thing. Run the crawl.

If you're dealing with the aftermath — traffic down, nobody sure when it started — the contact form is here. And for what to verify before, during, and after a migration: you moved the website but forgot the furniture.